Does Corporate Governance Reduce the Overinvestment of Free Cash Flow? Empirical Evidence from China

Transcription

1 Journal of Finance and Investment Analysis, vol. 2, no.3, 2013, ISSN: (print version), (online) Scienpress Ltd, 2013 Does Corporate Governance Reduce the Overinvestment of Free Cash Flow? Empirical Evidence from China Ji-fu Cai Abstract This paper theoretically and empirically investigates the relationship between corporate governance mechanisms and firm level overinvestment of free cash flow based on a broad cross-sectional sample of 1411 firm-year observations of listed companies in Shanghai and Shenzhen stock exchanges in China over the period 2003 to By following the creative approach to measure overinvestment and free cash flow suggested by Richardson (2006), the results show that there is a significantly posive association between overinvestment and free cash flow. Further analysis reveals that this posive association is mainly driven by state-owned enterprises sub-group. Finally, this paper explores the governance role of firm s board of directors and debt financing in controlling overinvestment of free cash flow. The results indicate that state-owned enterprises wh a large board of directors are more likely to engage in overinvestment of free cash flow. Contrary to the theoretical expectation, there is no evidence suggesting that the independence of the board of directors, as measured by eher the proportion of nonexecutive directors on the board or the seperation of roles of board chairman and CEO, is significantly negatively associated wh overinvestment of free cash flow. However, I find that both short-term debt (debt matury structure) and total leverage can significantly reduce the likelihood of state-owned enterprises overinvestment. Furthermore, the governance role of short-term debt in constraining overinvestment is even stronger for private-owned enterprises wh high free cash flow and low growth opportunies. However, bank loan has no impact on the reduction in the degree of overinvestment of free cash flow. These results above show that the board of directors is ineffective in alleviating the firm s level of overinvestment of free cash flow. According to relevant provisions of newly promulgated companies law in China, the board of directors should hold overall responsibily for ensuring that shareholders interests are not expropriated by self-interested managers, however, does t perform functions very well. School of Accounting, Jiangxi Universy of Finance and Economics, Nanchang, Jiangxi Province , China. Tel: Article Info: Received : July 4, Revised : July 27, Published online : August 15, 2013

2 98 Ji-fu Cai JEL classification numbers: G38, G21, G14 Keywords: Overinvestment; Free Cash Flow; Corporate Governance; Debt Financing 1 Introduction In a world wh perfect capal markets where there are no asymmetric information and transaction costs, and debt financing is free risk, Modigliani and Miller (1958) have confirmed that a company s investment decisions are independent of s financing decisions, and the market value of a company will be determined only by the future profabily and capal cost of s investment projects, which will achieve the maximum market value at the optimal level of investment. However, there are no condions that establish perfect capal markets in realy, Information asymmetries and transaction costs in the capal markets may give rise to agency conflicts and contract enforcement problems between shareholders and managers, which causes the actual investment expendures of a company deviating from s optimal level of investment, and thus results in the company s investment inefficient. Though the manifestations of inefficient investment include both overinvestment, which shows that a company may undertake some projects wh negative net present value, and underinvestment, which indicate that a company could forego or postpone some investment opportunies that would have posive net present value in the absence of adverse selection (Biddle, Hilary and Verdi, 2009), from a principal-agency perspective, if interest conflicts between the managers and shareholders are reflected in a firm s investment decisions, in order to obtain much more monetary and non-monetary private benefs associated wh a larger company size, such as pursu of power, as well as perquises, self-interested managers would continue to invest in some low-return or even loss projects that are beneficial from view point of managers but costly from the perspective of shareholders (Jensen, 1986). Overinvestment makes a company s funds troppo sunk in idle fields of production capacy, which wastes scarce resource and results in a reduction in company value. Therefore, in nature, overinvestment is not only whether a company s investment is efficient, but also an agency problem. The free cash flow hypothesis suggested by Jensen (1986) states that when there exist rich internal cash flows in excess of that required to fund all projects that have posive net present values which are discounted at the relevant capal cost, managers empire building incentives will create the potential to misuse those funds rather than pay them out to shareholders. In a model in which managers have an interest in growth, and can t credibly communicate their company s investment opportunies to the market, Stulz (1990) demonstrates that managers will tend to choose to invest too much when cash flow is high, and are forced to invest too ltle when cash flow is low. By influencing the resources under managers discretion, financing policies can reduce the costs of under- and overinvestment. Empirically, Richardson (2006) uses an accountingbased framework to measure overinvestment and free cash flow and finds that overinvestment is likely to occur in companies wh the highest levels of free cash flow, which provides the direct empirical support for Jensen s (1986) free cash flow hypothesis. At the same time, his evidence indicates that certain governance mechanisms, such as the presence of activist shareholders, appear to migate overinvestment. In the context of China, some scholars have also found that there exists overinvestment behavior abusing cash flows for Chinese listed companies. Fu (2011) examines the effect of overinvestment on the operating performance of SEO companies and finds that is managers

3 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 99 overinvestment incentives behavior that results in companies operating performance deterioration following seasoned equy offerings (SEOs). Based on the Jensen s free cash flow hypothesis, Liu (2006) and Li (2007) have studied the relationship between overinvestment and free cash flow, respectively, and found that a company s overinvestment is significantly posively associated wh s free cash flow. The financial objective of a company is to maximize s shareholders wealth or return on investment. Though overinvestment could increase the managers utily, as a behavior of abusing funds and destroying company value, overinvestment damages shareholders interests. Thus, when a company s free cash flow is high, how to force managers to return these surplus funds to shareholders rather than invest them in unprofable projects or waste them on organizational inefficiencies becomes an important problem that corporate governance mechanisms need to resolve. On the one hand, being a core instution arrangement of modern company governance structures, the board of directors performs two main functions of ratification and monoring for important decisions of a company including the investment activies (Fama and Jensen, 1983). While a large body of corporate governance lerature has investigated the effect of the board of directors on corporate agency problems and demonstrated an efficient board of directors can play a major role in reducing agency conflicts between shareholders and managers (Yermack, 1996; Peasnell, Pope and Young, 2005; Cai, 2007), relatively few papers have directly explored the relationship between overinvestment of free cash flow and the board of directors. A potentially more important question is whether the board of directors can control the overinvestment of free cash flow and improve the efficiency of resource allocation of a company, and protect the outside shareholders interests against expropriation imposed by self-interested managers. On the other hand, Jensen (1986) and Stulz (1990), among others, argue that a potential way to resolve the overinvestment problem of free cash flow is debt creation. Due to s hard constraint attribute of payment of principals and interests, debt reduces the cash flows available for value-destroying empire-building projects at the discretion of managers, and thus effectively lower the agency costs of free cash flow. The effects of the supervision of managers and improvement of organizational efficiency of debt are called as the control hypothesis for debt creation (Jensen, 1986). Utilizing the basic principle of control hypothesis for debt creation, Gul and Tsui (1998), Jaggi and Gul (1999), Wang (2004) test the role of debt in reducing the agency costs of free cash flow, respectively. Gul and Tsui, Jaggi and Gul both find that debt can significantly reduce the free cash flow at the discretion of managers and thus provide the direct support for Jensen s control hypothesis for debt creation. On the contrary, using Chinese listed companies firm-level observations as research sample, Wang does not find debt has the governance role of reducing the agency costs of free cash flow. While Lang, Ofek and Stulz (1996), and Aivazian, Ge and Qiu (2005a) directly explore the influence of debt on firm investment and document that debt is significantly negatively associated wh investment. In another paper, Aivazian, Ge and Qiu (2005b) directly test the correlation between matury structure of a firm s debt and s investment decisions, and find that, after controlling for the effect of the overall level of leverage, a higher percentage of long-term debt in total debt significantly reduces investment for firms wh high growth opportunies. In contrast, the relationship between debt matury structure and investment is not significant for firms wh low growth opportunies. Through systematical analysis of these lerature, I find that, most studies mainly focus on how debt reduce the agency costs of free cash flow. By contrast, few papers have directly

4 100 Ji-fu Cai studied the governance role of debt in controlling the overinvestment of free cash flow. Furthermore, when a majory of the scholars study the control hypothesis for debt creation, they usually consider classes of debt as homogeneous, and few have paid attention to the differences in the governance effect of classes of debt wh different seniories. In fact, Since there is a huge differences for corporate debt in terms of matury, sources and priories, different debt financing contract arrangements inevably lead to the differences in governance efficiency of classes of debt. Therefore, when exploring control hypothesis for debt creation, may arrive at a confused or wrong research conclusions whout considering the difference in governance efficiency of classes of debt. In sum, based on the analysis above, I think the evidence on the possible effect of governance mechanisms on overinvestment of free cash flow is scarce. The purposes of this paper seek to address the following questions: (1) Under the special instutional background of a transional economy, such as China, whether there is overinvestment of free cash flow in Chinese listed companies. (2) If Chinese listed companies tend to engage in overinvestment when free cash flow at managers discretion is high, whether both the board of directors and debt can reduce the overinvestment of free cash flow and relieve the agency conflicts between shareholders and managers in terms of the use of surplus funds. The answer to the first question constutes the base for further studying the second question. The principal tests of this paper suggest that state-controlled companies wh high free cash flow are more likely to engage in overinvestment, but the posive relationship between overinvestment and free cash flow does not exist in provate-controlled companies. I also find that state-controlled companies wh a large board of directors tend to use their free cash flow in overinvestment, Contrary to the theoretical expectation, the results of this paper do not provide evidence that the independence of the board, as measured by eher the the proportion of non-executive directors on the board or the seperation of roles of board chairman and CEO can significantly constrain the overinvestment of free cash flow. Both short-term debt and total debt are significantly negatively associated wh overinvestment. Furthermore, short-term debt can play an even more important roles in reducing the overinvestment of private-controlled companies wh high free cash and low growth opportunies. However, the governance role of private debt, such as bank loan, in controlling a firm s overinvestment of free cash flow is weakened, and there is no evidence suggesting that bank loan can significantly reduce the overinvestment of free cash flow relative to other debt sources. These research results above have important policy implications due to the heightened interest in corporate governance matters from governments and regulators (Davidson, Goodwin-Stewart and Kent, 2005). The remainder of this paper is organized as follows. Section 2 presents the theoretical analysis and related hypotheses. In section 3, I provide a brief description of the sample selection, the variable definions and methodology specification. It also discusses the measurements of overinvestment and free cash flow by following the investment expectation model creatively suggested by Richardson (2006). The main results are reported in section 4. The final section summarizes findings of this paper and discusses some policy implications.

5 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? Instutional Background, Theoretical Analysis and Research Hypotheses Since free cash flow hypothesis advanced by Jensen (1986), free cash flow has become one of the most important factors to be considered in the overinvestment research. According to Jensen s (1986) definion, free cash flow is cash flow in excess of that required to fund all projects that have posive net present values when discounted at the relevant cost of capal (Jensen, 1986). Due to agency problems, there are interest conflicts in terms of the use of free cash flow between managers and shareholders. Theoretically, if the firm had excess cash beyond that needed to fund available posive NPV projects (including options on future investment), from the perspective of increasing shareholders wealth, would distribute free cash flow to shareholders in the form of extra dividends. However, returning free cash flow to shareholders will reduce resources under control of managers which could be used to build empires to increase their personal utily. Thus, managers have incentives to hoard and abuse free cash flow, and invest the excess funds in some projects wh negative NPV which are beneficial from managers perspective but costly from shareholders perspective. Through continuously investing in negative NPV projects, managers can not only control more resources and acquire more persquis consumption, but also upgrade their powers in the firm. Especially for those firms whose free cash flow is high (i.e., free cash flow is posive), but growth prospects are poor, the incentives for managers to undertake overinvestment are usually even more severe. therefore, free cash flow hypothesis holds that firms wh large free cash flow are more likely to engag in overinvestment. These overinvestments, though enhancing managers private benefs, destroy company value, and thus reduce shareholders wealth. Richardson (2006) finds that overinvestment is mainly concentrated in firms wh highest levels of free cash flow. On the contrary, overinvestment is less likely to occur in firms wh low free cash flow. Based on the analysis above, I can put forward the first hypothesis: H1: Overinvestment is significantly posively associated wh free cash flow. During the process of economic transion in China, in order to satisfy the fund demand of state-owned enterprises to realize the sustainable growth of inner-system economies, Chinese government adopts ultrastrong financial control policy characterized by financial repression and ownership discrimination. Financial repression and ownership discrimination under ultrastrong financial control policy have resulted in the abily of private-controlled enterprises to raise external funds generally weaker than that of statecontrolled enterprises, which lead to private-controlled enterprises facing much more severely financing constraints in the capal markets. In order to obtain the funds required for investment externally, private-controlled enterprises are usually forced to pay very high cost premium for external financing. Therefore, private-controlled enterprises managers have even stronger incentives to use company s funds effectively than those of state-owned enterprises. In other words, though the investment expendures of privatecontrolled enterprises may be also distorted and inefficient due to information asymmetry in the capal markets and agency conflicts, the inefficient degree of investment of private-controlled enterprises is significantly lower than that of state-owned enterprises, and the forms of inefficient investment of private-controlled enterprises are more likely to arise from underinvestment rather than overinvestment. On the contrary, because of the policy loans of the state-owned banks and the expectation of soft budget constraint bailing out from the governments at all levels when falling into financial distress, as well as

6 102 Ji-fu Cai instutional arrangements of the stock markets sevicing for the difficulties of the stateowned enterprises in China, state-owned enterprises generally have a much higher abily to obtain external financing and thus face lower financing constraint than that of privatecontrolled enterprises. As a result, managers of state-owned enterprises are more prone to generate self-interested behaviors, and less likely to use the company s funds effectively, and thus giving rise to even higher agency problems between shareholders and managers. In order to obtain much more monetary and non-monetary benefs associated wh a large company size, managers of state-owned enterprises have more strong incentives to engage in overinvestment, which inevably reduces the company s investment efficiency (Cai, 2012). Thus, according to the theoretical analysis above, my second hypothesis could be ststed as follows: H2: The overinvestment problem of free cash flow is even more severe in state-owned enterprises than in private-controllrd enterprises in China, other things being equal. Though the theoretical analysis above shows that, due to agency problems between managers and outside shareholders, firm tends to engage in overinvestment when s free cash flow is high. However, both free cash flow and overinvestment are associated wh the optimal level of investment. According to neoclassical investment theory (Jorgenson, 1971), since the investment demand of a company is an increasing function of s investment opportunies, a company s optimal level of investment will systematically vary wh s qualy of growth prospects. To be exact, the optimal investment level will be higher for those firms that have abundant valuable growth opportunies in the future than that of other firms wh poor growth opportunies. Thus, through influencing the optimal level of investment, growth opportunies will impose an important effect on the level of free cash flow and overinvestment, which results in the relation between free cash flow and overinvestment potentially changing wh growth opportunies of a company. The probabily of overinvestment is much higher for companies wh serious agency costs of free cash flow, which are lack of good investment opportunies and simultaneously hold large excess cash flows. By developing a measure of free cash flow using Tobin s q to distinguish between companies that have good good investment opportunies and those that don t, and using a sample of successful tender offers, Lang, Stutz and Walking (1991) find that companies wh high cash flow and low growth opportunies are more likely to engage in acquision activies that result in a reduction in company value. On the contrary, the probabily of overinvestment is least for companies wh low cash flow and high growth opportunies. accordingly, I can put forward the third hypothesis below: H3: If free cash flow hypothesis is ture, the overinvestment problem is much stronger for companies wh low growth opportunies and high free cash flow compared wh other types of companies. Diffuse shareholders constrained by collective action problems must rely on the board of directors to monor and deter managers from implementing policies that diverge from shareholders interests (Hanson and Song, 2006). In modern company, the board of directors is widely believed to play an important role in corporate governance, particulary in constraining and ensuring that managers act in the interests of outside shareholders (Fama and Jensen, 1983) and regard the board of directors as one of the most important control mechanisma available since constutes the apex of a company s internal governance structures. Because the board of directors links shareholders who provide capal and managers who use these capal to create value together, is regarded as the core of modern corporate governance mechnisms in some corporate governance lerature. From an agency perspective, the abily of the board of directors to act as an effective

7 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 103 control mechanism is often dependent upon s size and independence from the managers (Beasley, 1996; Davidson, Goodwin-Stewart and Kent, 2005). The role of board size in corporate governance has not yet reached agreement among scholars. Some scholars agrue that if board size is too large, will result in the directors acting slowly and divergent views, and thus lead to the directors powers and responsibilies unclear and weaken their monoring functions, which causes a company operation inefficient. Jensen (1993) argues that, due to the inherent coordination problems inside the board of directors together wh managerial discretion, the operating efficiency is much higher for a company wh small size board of directors than that wh large size board of directors. When the number of directors on the board is beyond (or exceeds) seven or nine persons, the likelihood of the board of directors effectively monoring managers is smaller and more likely to be controlled by managers. Yermack (1996) argues that companies wh smaller board tend to be more effective in monoring managers behavior. As evidence, Newell and Willson (2002) have documented that the most efficient board of directors should be comprised of 5-9 members. However, other scholars believe that a large board of directors is much more beneficial to corproate governance than a small board of directors. More (2002) argues that the board of directors is too small to meet the demand for range of management, and that a board of directors wh nine to fifteen members is the most suable. Given the board size, the governance effectiveness of the board of directors is usually determined by s independence from management. Independence of the board of directors refers to the extent to which a board of directors is comprised of non-executive directors who have no relationship wh the firm beyond the roles of directors and whether there is a separation of the roles of board chairman and the chief executive officer (Davidson, Goodwin-Stewart and Kent, 2005). Directors independence as well as their power and willingness to constrain managers malevolent behavior plays a crucial role in the success of governance structure of a company (Hanson and Song, 2006). The governance lerature stresses the the role of non- executive directors in resolving agency problems between managers and shareholders through the creation of appropriate employment contracts and the subsequent monoring of managerial behavior (Peasnell, Pope and Young, 2005). Fama and Jensen (1983) argue that the incentives to maintain the value of their reputation capal in the external labor markets enable non-executive directors to be less likely to collude wh managers to expropriate shareholders wealth, which thus ameliorates the interest conflicts between managers and shareholders. Therefore, a non-executive director who is entirely independent from management is expected to provide the greatest protection in monoring managers actions and ensuring that managers are pursuing policies consistent wh shareholders interests (Baysinger and Butler, 1985). It is generally accepted that a higher proportion of non-executive directors on the board is more effective in constraining managers discretion. Consistent wh view above, Rosenstein and Wyatt (1990), Byrd and Hickman (1992) find that the proportion of non-executive directors on the board is posively associated wh a company performance. Lin, Pope and Young (2003) show that stock prices of a company respond posively to the announcement of appointment of non-executive directors, namely, the appointment of non-executive directors increases the company value. McWilliams and Sen (1997) documents that board effectiveness in protecting shareholders wealth is an incresing function of the proportion of non-executive directors on the board. Another independence of the board of directors relates to dualy, which occurs when the same person undertakes the combined roles of chief executive officer and board chairman

8 104 Ji-fu Cai (Hanson and Song, 2006). According to principal-agency theory, in the modern company where ownership separates from management, agency problems are mainly reflected in the interest conflicts between managers and shareholders. Using the board of directors to monor managers is an important governance mechanism that shareholders safeguard their own interests from expropriation by managers. The abily of the board of the directors to fulfil monoring function will be weakened when chief executive officer also serves as board chairman. The appointment of the chief executive officer to the posion of board chairman can lead to a concentration of power (Beasley, 1996) and possible conflicts of interest, thus resulting in a reduction in the level of monoring (Davidson, Goodwin-Stewart and Kent, 2005). As a specific form of agency problem in the modern company, the occurrence and level of overinvestment of free cash flow is obviously influenced by the governance effectiveness of the board of directors. Based on the analysis above, I can put forward these three hypotheses as follows: H4a: Overinvestment of free cash flow is posively associated wh the board size. H4b: Overinvestment of free cash flow is negatively associated wh the board size. H5: Overinvestment of free cash flow is negatively associated wh the proportion of nonexecutive directors on the board. H6: Overinvestment of free cash flow is negatively associated wh the separation of the roles of board chairman and CEO. a company s overinvestment problem is generally associated wh s free cash flow, as well as the information asymmetry in the capal markets. Previous studies have showed, as a control mechanism, debt has the functions of migating managers moral hazard and lowering agency costs of free cash flow by forcing managers to disgorge surplus funds, and therefore reducing the amount of cash under their discretion. Debt is thus hypothesized to be able to alleviate the incentives for managers to overinvest in negative net present value projects which reduce shareholders wealth. However, when scholars analyze the governance role of debt, they always consider classes of debt of a company raising from different sources as homogeneous. In fact, each debt is different in terms of matury, source and priory. Different debt financing contract arrangements will lead to the differences in governance efficiency of debt. First, as for debt matury structure, Although, due to s role of hard claims in constraining managers, both short-term debt and long-term debt would play an important role in the modern corporate governance, the characteristic or emphasis of restricting managers behavior is different between shortterm debt and long-term debt (Yang and Zheng, 2004). Several prior studies have showed that the governance role of short-term deb is mainly reflected in the liquidation of a company and restriction on managerial discretion over free cash flow (Hart and Moore, 1995) 12, and delivering high qualy signal to the outside investors (Flannery, 1986). On the contrary, the governance role of long-term debt is mainly focused on preventing the company from raising new capal for unprofable investment against future earnings from assets in place (Hart and Moore, 1995). Therefore, relative to the long-term debt, short-term debt has an advantage in reducing the agency costs associated wh free cash flow and information asymmetry in the capal markets. Second, as far as debt sources are 2 Hart and Moore (1995) have theoretically showed that nonpostponable, short-term debt could force managers disgorge cash flows that might otherwise be used to make unprofable but empirebuilding investment, and trigger liquidation in states of the world where the firm s assets are more valuable elsewhere.

9 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 105 concerned (wh respect to debt sources), private debt, such as bank loans, will be more efficient in curbing and monoring the managerial behavior than public debt, whose typical form is a company s publicly issued bond, due to the following reasons: The first is that, as a professional lending instutions, banks can timely grasp internal information that a company has yielded yet not disclosed externally through using the companies loan applications and their bank accounts. Thus, banks have advantages in collecting and processing information about companies. The second is that, because of their larger loan amount and longer loan term, in most cases, banks are the main representative of credors participating in corporate governance and have the abily to intervene wh the companies activies when is necessary. The third is that loan renegotiation between bank and company can transfer signals on the company s qualy to capal markets, which thus reduces information asymmetry between company and outside investors. Therefore, as a credor, bank has a comparative advantage in participating in corporate governance and monoring managers of companies. Taken together, the matury structure and sources of debt constute two important corporate governance mechanisms. In this paper, I employ the ratio of current liabilies to total debt and the ratio of bank loan (the sum of short-term loan and long-term loan) to total debt (hereforth referred to as the proportion of short-term debt and bank loan ratio, respectively) to proxy for two corporate governance mode of debt, which are used to reflect differences in goverance role of matury structure and sources of debt in controlling agency problems arising from overinvestment. In addion, I also use the total debt-to-asset ratio (leverage) to proxy for credors incentives to monor managers. Generally speaking, as debt increases, the default risk of a company also increases. Credor, therefore, have even higher incentives to monor managers. Based on the analysis above, this can lead to the following hypotheses: H7: Overinvestment of free cash flow is negatively associated wh the ratio of short-term debt. H8: Overinvestment of free cash flow is negatively associated wh the ratio of bank loan. H9: Overinvestment of free cash flow is negatively associated wh total debt-to-asset ratio. A series of research results have confirmed that agency problems associated wh underinvestment and overinvestment systematically vary wh a company s growth opportunies and exhib significant variation (Myers, 1977; Jensen, 1986). The companies wh high growth opportunies will face much more underinvestment problem (Myers, 1977), however, companies wh low growth opportunies are often suffered from an overinvestment problem (Jensen, 1986). Since the agency problems faced by the companies wh high growth opportunies are fundamentally different from those encountered by the companies wh low growth opportunies, the control effect of debt varies wh the company s growth opportunies. In particular, because companies wh low growth opportunies and high free cash flow are more likely to give rise to agency problems of overinvestment, the control hypothesis implies that the governance function of debt is more important in companies that generate large cash flows yet have poor growth prospects, and even more important in companies which are falling into recession. In these companies, the potential pressure to squander cash flows by investing them in unprofable projects is the most serious (Jensen, 1986). On the contrary, wh respect to rapidly growing companies wh large and highly profable investment projects but low free cash flow, since such companies will be forced to resort to capal markets to raise funds regularly, the markets have an opportuny to evaluate these companies, their

10 106 Ji-fu Cai managers, and their proposed projects. In this case, other governance mechanisms, such as bankers and analyst followings, will play a crucial role in monoring managers behaviors, which means that the governance function of debt will not be important for this kind of companies (Jensen, 1986). Some studies have empirically confirmed that there is an interaction between the corporate debt and growth opportunies. McConell and Servaes (1995) find that the relationship between corporate value and debt is negatively correlated for the rapidly growing companies, however, posively related for companies wh low growth prospects. According to the theoretical analysis above, the following hypothesis can be proposed: H10: The governance function of debt controlling overinvestment of free cash flow is significantly affected by the company s growth opportunies. It is very important to note that s hard constraint attribute and the existence of the efficient bankruptcy instutions constute an essential prerequise of the achievement of control effect of debt. As for China, there are many deficiencies in these aspects, which are focused on that soft budget constraint problems of debt resulting from governments at all levels administrative intervention in bank loan decision and irrationaly of debt financing of state-owned enterprises have not been fundamentally changed so far. Consequently, whether debt can really play a role in reducing and controlling the overinvestment of free cash flow of Chinese listed companies is still an important problem which should be tested by using normative empirical research approaches. 3 Sample Selection and Research Design 3.1 A Framework to Measure the Overinvestment and Free Cash Flow In order to construct mearsures of underinvestment and overinvestment, I follow the approach suggested by Richardson (2006) and first estimate a model that predicts expected investment of a company and then use residuals from this model as a proxy for inefficient investment. The model that has been modified is as follows: I Gr 0 1 LnAge 7 1 Cash 2 1 LnTA Ind Year 3 1 Roa 4 1 Lev 5 1 I 6 1 (1) Where i is the sample company and t indicates the year in the sample period, respectively; I is the firm s capal expendures and measured as cash paid to acquire fixed assets, intangible assets and other long term assets minus net cash received from the sale of fixed assets, intangible assets and other long term assets in period t scaled by the average book value of total assets as of the end of year t-1 and t. The prior period s firm-level (lagged) investment is included in model to capture non-modeled firm characteristics that affect investing decisions (Richardson, 2006) and the acceleration effect of investment. Gr is the firm s investment opportunies as of year t-1. In empirical studies, the variables commonly used to measure the company s investment opportunies are Tobin q and sales growth, repectively. Tobin q is defined as the ratio of the market value of the company s assets to their replacement cost at the start of the fiscal year. The market value of the company is the sum of the market value of the equy, the value of short term debt and the value of long term debt. The replacement cost of assets is proxied by the book value of

11 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 107 total assets. Tobin q is an imperfect measure of investment opportunies because is an average value rather than marginal value (Hayashi, 1982; Lang, Stulz and Walking, 1991). Further, marginal q self is difficult to measure and Tobin q calculation will use stock prices. Due to the inefficiency and functional fixation problems of stock markets in China, employing Tobin q to proxy the company s investment opportunies will bring measurement errors. In addion, Alti (2003) has also showed that, since Tobin q mainly reflects option value relating to firm long term growth potential but doesn t provide information about investment opportunies in the near-term, Tobin q performs as a noisy measure of short-term investment expectations. Thus, to control possible measurement error in Tobin q as a proxy for investment opportunies, I use sales growth as a proxy for a company s investment opportunies to estimate the regression. Cash is the firm s cash and cash equivalent divided by the book value of total assets as of year t-1. LnTAis the natural logarhm of book value of total assets as of year t-1, used to control the effect of company size on the investment. Roa is return on assets as of year t-1, equal to the ratio of the prof before interest and tax to the book value of total assets. Prior period s returns are included as an addional variable to capture growth opportunies not reflected in Gr. Lev is debt-to-asset ratio and measured as the book value of total debt (the sum of shortterm debt and long-term debt) divided by the book value of total assets as of year t-1. LnAge is the natural logarhm of the number of years the firm has been listed on the stock exchanges in China since IPO. Firm level investment is lessed when is more difficult to raise addional cash to finance the new investment as captured by leverage, firm size, firm matury and level of cash (Richardson, 2006). Finally, I include industry indicators, Ind, and year indicators, Year, since firm level investment patterns may systematically vary wh differences in industry and are affected by fluctuation in macro economic condions. For the purpose of industry classification, the Standard Industry Classification Code of China Securies Regulatory Commission (CSRC) is adopted. According to Standard Industry Classification Code of China Securies Regulatory Commission (CSRC), I constructed 20 industry dummy variables, consistent wh prior research, such as Xia and Fang (2005). is error term. The fted values from the regression model (1) is the estimate of the expected level of investment, EI. The unexplained portion (or the error term) is the estimate of the unexpected investment, UI, which reflects the degree of a company s investment distortion. I measure investment efficiency using the residual from the model (1). If the residual is greater than 0, indicates that firm is overinvesting. On the contrary, if the residual is less than 0, means that firm is underinvesting. Both overinvestment and underinvestment are decreasing in investment efficiency (Biddle, Hilary and Verdi, 2009). Free cash flow can be defined as cash flow beyond what is necessary to maintain assets in place and to finance expected new investment (Richardson, 2006). According to the definion above, after calculating firm s expected investment for a particular firm, free cash flow can be computed as the difference between the firm s net cash flows from operation and s expected level of investment (EI), as estimated wh regression model (1), and thus obtained as follows: FCF OCF EI (2)

12 108 Ji-fu Cai Where FCF, OCF and EI is the firm s free cash flow, net cash flows from operating activies and the expected level of investment in period t of a company and normalized by the average book value of total assets as of the end of year t-1 and t, repectively. 3.2 Sample Selection and Data Sources For the study of this paper, the inial sample are selected from all non-financial companies listed on Shanghai or Shenzhen stock exchanges in China during the period 2003 to To ensure the validy of the data collected and simultaneously minimize the effect of other factors on the research results, I exclude from our inial sample those companies whose main operational business has ever experienced substantial change. Also excluded are firms which have extreme outliers and those whose financial information is seriously inadequate or obviously misrecorded. At the same time, the privatized companies whose controlling private ownership came into being through the block transfer of state shares after IPO are also excluded. After these exclusion are made, I then obtain a pooled sample wh 7215 firm-year observations in total over 8 years. On this basis, I use the model (1) to regress these 7215 firm-year observations. According to the study purpose of this paper, I confine the analysis to the subgrouds of companies whose unexpected investment and free cash flow are both greater than zero (posive). Finally, the sample is left 1411 firm-year observations. Eher financial data or nonfinancial data used in this paper, such as investment expendures, growth opportunies, return on assets, the book value of asset and equy, short term debt, bank loan, debt-toasset ratio (total leverage), the composion or characteristic of the board of directors, age (the number of years listed on stock exchanges since IPO), and company s ownership identy et al., are all obtained from disclosure made in annual report of listed companies published by Shanghai Wind Information Co., Ltd. of China, a leading Bloomberg-style data provider in China, and the China Securies Markets and Accounting Research (CSMAR) database prepared by Shenzhen GTA Information Technology Limed Company, another major data provider in China. Table 1 presents the characteristics of full sample and subsamples of state-controlled companies and private-controlled companies by year and industry. It is evident from the year distribution outlined Panel A that, among 1411 firm-year observations, statecontrolled companies and private-controlled companies account for 1135 and 276 of observations in my sample, respectively. Moreover, in each year, the observations of statecontrolled companies are all more than those of private-controlled companies, indicating that the probabily of engaging in overinvestment is more likely to occur in statecontrolled companies rather than private-controlled companies. Nevertheless, the number of overinvestment of private-controlled companies increase steadily from 14 in 2003 to 69 in 2010, suggesting that, wh the increasing number of listed companies controlled by private enties, the likelihood of private-controlled companies to engage in overinvestment is also gradually increase. Panel B reports the industry distribution of full sample and subsamples of state-controlled companies and private-controlled companies. As wh Panel A, in each industry, state-controlled companies all exhib a higher possibily of undertaking overinvestment squandering funds. However, private-controlled companies operating in the machinery and equipment, medical and biological product sectors tend to have a higher number of overinvestment at 53 and 38, respectively.

13 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 109 Table 1: Distribution of Sample Panel A: By year This panel outlines the distribution of full sample and subsamples of state-controlled companies and private-controlled companies by year. Year Full sample State-controlled companies Private-controlled companies Total Panel B: By industry This panel outlines the distribution of full sample and subsamples of state-controlled companies and private-controlled companies by industry. Industry Full sample State-controlled companies Privatecontrolled companies Agriculture Communication Conglomerate Construction Culture, Sport and Entertainment Electric, Gas, and Water Electron Textile and Clothing Machinery and Equipment Metal and Nonmetal Lumber and Furnure Other Manufacturing Petroleum and Chemical Food and Beverage Medical and Biological Products Papermaking and Printing Mining Public Utily Real Estate Transportation Wholes and Retail trade Total

14 110 Ji-fu Cai 3.3 Model Specification and Variable Definions The basic regression specifications used to test the hypotheses developed in this paper take the following two forms. OverI 0 1FCF 2Gr 3Roa 1 4LnAge Ind Year (3) OverI FCF Gr FCF * Gr Board Dir Dul Board * FCF Dir * FCF Dul * FCF ShortDebt Bank Lev ShortDebt * Gr Bank * Gr Lev * Gr ShortDebt * Gr * FCF Bank * Gr * FCF Lev * Gr * FCF Roa LnAge Ind Year (4) In the model (3) and (4), i and t are firm and year indicators, respectively; OverI is the posive residuals estimated from regression model (1), which is used as a proxy for the firm s level of overinvestment. FCF is free cash flow that a company holds and measured as the difference between net cash flows from operating activies and the expected level of investment estimated from regression model (1) scaled by average book value of total assets as of the end of year t-1 and t. Gr is the firm s sales growth as of the end of year t, equal to the change in sale revenues between current year and previous year divided by total sale revenues of previous year, indicating a firm s investment opportunies. In model (4), The interaction term FCF Gr is used to examine the effect of investment opportunies on the relation between free cash flow and overinvestment. Based on hypothesis 2, I expect the coefficient of FCF Gr should be significantly negative. Board is the board size, as measured by the total number of directors on the board. Based on earlier studies, I conjecture that a smaller board will play a stronger monoring role. Dir is the proportion of non-executive directors on the board. Dul is a dummy variable taking the value of 1 if the roles of chief executive officer and board chairman are overlapped, and 0 otherwise. The interaction terms, Board * FCF, Dir * FCF and Dul * FCF, are used to explore the effect of the size and the independence of the board on the relation between free cash flow and overinvestment, respectively, namely, whether an effective board of directors can significantly reduce a firm s overinvestment of free cash flow. ShortDebt is year-end total current liabilies divided by year-end total debt, reflecting a firm s debt matury structure. Bank is bank loans acquired from banks scaled by total debt, which is used to measure debt sources. Lev is the sum of the year-end book value of short-term debt and long term-debt deflated by the year-end book value of total assets. Jensen (1986) argues that debt can serve as a device liming managers discretion and reduce the company s free cash flow, and thus forcing managers to reduce overinvestment (debt control hypothesis). Jaggl and Gul (1999) find that there is a significantly posive association between free cash flow and debt for firms wh low investment opportuny set, which provide direct support to Jensen s (1986) debt control hypothesis. The interaction terms, ShortDebt * FCF, Bank * FCF and Lev* FCF, are employed to test Jensen s control hypothesis for debt creation as well as the differences in governance effect among debt wh different matury and sources. If debt financing can effectively reduce overinvestment of free cash flow, the coefficients of ShortDebt * FCF, Bank * FCF and Lev* FCF are expected

15 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 111 to be all significantly negative. The tripple interaction terms, ShortDebt * FCF * Gr, Bank * FCF * Gr and Lev * FCF * Gr are used to examine the hypothsis 10 whether the governance function of debt financing controlling overinvestment of free cash flow is affected by growth opportunies. If the coefficients of ShortDebt * FCF * Gr, Bank * FCF * Gr and Lev * FCF * Gr are all signficantly posive, then the hypothsis 10 is supported empirically. Remaining variables, such as Roa and LnAge, are all as previously defined. In addion, in order to deeply investigate the difference in overinvestment of free cash flow between state-controlled companies and private-controlled companies, and explore how ownership identy of a company influences the role of governance mechanisms in controlling the overinvestment of free cash flow, at the same time avoid the multicollineary between variables used in this paper, I further separate listed companies into state- and private-controlled subsamples based on the identy of the company s ultimate controlling shareholder and explore the effect of corporate governance on the two subsamples separately. When the company s ultimate controlling shareholder is the governments at all levels, such as the bureaus of state assets management, finance bureaus and bureaus in charge of different industries or other government agencies et al., I regard as a state-controlled company. On the contrary, if the company s ultimate controlling shareholder is non-government uns, such as entrepreneurs, townships and villages, and foreign companies, is correspondingly treated as a private-controlled company, and reestimate the regression model (3) and (4), respectively. Ultimate controlling shareholder of a company is identified through reviewing s published annual report. 4 Results 4.1 Analysis of Investment Model Table 2 provides the descriptive statistics for the variables used to estimate the investment model (1). The mean (median) company in the sample engages in investment activies equal to (0.049) of average total assets as of the end of year t-1 and t and has an average (median) Gr equal to (0.147) during the sample period. The mean (median) cash across all firm-years equals to (0.138). The mean (median) value for the firm operating performance is (0.052), indicating that firms performed poorly during sample period on the whole and some firms have suffered from an even more serious loss (the minimum value of the firm operating performance is ). The average (median) company has reported debt-to-asset ratio of (0.482), the maximum debt-to-asset ratio is 0.996, indicating that some companies have fallen into serious financial distress during the period of study. On average, company has been listing 7.60 years on the stock exchanges in China after IPO. The sample period for investment expectation model (1) is For each variable, I report the number of firm-year observations, mean, median, minimum (Min), maximum (Max) and standard error (Std), where I is the firm s investment expendures and measured as cash paid to acquire fixed assets, intangible assets and other long term assets minus net cash received from the sale of fixed assets, intangible assets and other long term assets in period t scaled by average book value of total assets as of the end of year t- 1 and t. Gr is the firm s investment opportunies and as measured by sales growth.

16 112 Ji-fu Cai Cash is the firm s cash and cash equivalent divided by the book value of total assets as of year t-1. LnTAis the natural logarhm of the book value of total assets as of year t-1, used to control the effect of the size of a company on s investment. Roa is return on assets as of year t-1, equal to the ratio of the prof before interest and tax to the book value of total assets. Prior period s profabily is included as an addional variable to capture growth opportunies not reflected in Gr. Lev is debt-to-asset ratio and equal to the book value of total debt (the sum of short-term debt and long-term debt) divided by the book value of total assets as of year t-1. Age is the number of years a company has been listed on the stock exchanges in China since IPO. Table 2: Descriptive Statistics for the investment expectation model (1) Variable No. of obs Mean Median Min Max Std I Gr Cash LnTA Roa Lev Age Table 3 presents the regression results for the investment expectation model (1) based on the data of 7215 firm-year observations during the period , in which the dependent variable is the firm s capal expendures. This model is used to determine the expected investment level and overinvestment of a company. The expected level of investment is the fted values ( EI ) and overinvestment (OverI ) is the posive residuals estimated from the model (1). The model of investment expendures in the column (1) of Table 3 only includes investment opportunies which are proxied by sales growth in period t-1 and industry and annual fixed effects as independent variables. The coefficient on Gr is and significantly posive at 1 percent level, indicating that investment demand is an increasing function of growth opportunies, and this model explains 12.9% of the variation in investment expendure. The model of investment expendures in the column (2) of Table 3 that includes all control variables, such as cash balance, company size, debt ratio, the natural logarhm of the number of years listed on the stock exchanges in China, operating performance and prior investment expendures, explains 36.7% of the variation in investment expendures. However, when I include growth opportunies and all other control variables together to regress the model of investment expendures in the column (3) of Table 3 (model (3)), doesn t significantly increase explanatory power (the adjusted R-square of model (3) is 36.8%) and the coefficient on Gr has become much smaller, though the signs of all variables are the same as predicted. Nevertheless, in subsequent analysis I still rely on the model (3) in Table 3 as the baseline to decompose investment expendures of a company into expected investment and unexpected investment. This table provides the regression results for model (1): I Gr 0 1 Cash LnTA Roa Lev I LnAge Ind Year

17 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 113 where I is the firm s investment expendures and measured as cash paid to acquire fixed assets, intangible assets and other long term assets minus net cash received from the sale of fixed assets, intangible assets and other long term assets in period t scaled by average book value of total assets as of the end of year t-1 and t. Gr is the firm s investment opportunies and as measured by sales growth. Cash is the firm s cash and cash equivalent divided by the book value of total assets as of year t-1. LnTAis the natural logarhm of book value of total assets as of year t-1, used to control the effect of the size of company on the investment. Roa is return on assets as of year t-1, equal to the ratio of the prof before interest and tax to the book value of total assets. Prior period s profabily is included as an addional variable to capture growth opportunies not reflected in Gr. Lev is debt-to-asset ratio and equals total debt (the sum of short-term debt and long-term debt) divided by the total assets as of year t-1. LnAge is the natural logarhm of the number of years the company has been listed on the stock exchanges in China since IPO. Ind and Year are a vector of industry and year indicator variables, repectively, which are used to capture year and industry fixed effect. According to Standard Industry Classification Code of China Securies Regulatory Commission (CSRC), there are 20 industry dummy variables in the regression. is error term. Industry and year fixed effect are controlled for but not reported for the sake of space. T- statistics are presented below the estimated coefficients; ***, **, * indicate two-tailed statistical significance at the 1%, 5%, and 10% level, respectively. Table 3: The multivariate regression results of the investment model (1): Variable Predicted Model sign (1) (2) (3) Intercept? 0.045*** Gr *** *** Cash *** *** LnTA *** *** Roa *** *** Debt ** *** I *** 0.437*** LnAge *** *** Ind Included Included Included Year Included Included Included 2 AdjR F *** *** *** No. of obs

18 114 Ji-fu Cai 4.2 Analysis of Corporate Governance and Overinvestment of Free Cash Flow Descriptive statistics and univariate test Table 4 provides descriptive statistic information on the main variables used to estimate model (3) and (4). By construction, there are 1411 companies wh posive free cash flow classified as overinvesting companies. The mean (median) overinvestment across all firmyears equals to (0.028) of average book value total assets as of the end of year t-1 and t, and the standard deviation of the overinvestment is 0.061, indicating that there is a variation in overinvestment among companies during sample period. The mean (median) company in the sample has a FCF of (0.046), suggesting that majory of companies hold surplus funds. The mean (median) value for the investment opportunies is (0.181). The average (median) company has about 9.55 (9) directors on the board and the mean (median) value for the proportion of non-executive directors on the board is (0.333). Some 86 percent of companies maintain a separation of the roles of the chief executive officer and board chairman. The mean (median) values for short-term debt, bank loans and total leverage are respectively (0.903), (0.407) and (0.494), wh standard deviation of 0.174, 0.237, and 0.181, indicating that, on average, companies hold more short-term debt than long-term debt. This table presents descriptive statistics for the regression variables used in the hypothesis tests to examine the relationship between free cash free and overinvestment as well as the effect of corporate governance on the relationship between free cash flow and overinvestment for a sample of 1411 firm-year observations over the period For each variable, I report the number of firm-year observations, mean, median, minimum (Min), maximum (Max) and standard error (Std), where OverI is the posive residuals estimated from regression model (1), which is used as a proxy for the firm s level of overinvestment. FCF is the level of free cash flow that a company holds and measured as the difference between net cash flows from operating activies and the expected level of investment estimated from regression model (1) scaled by average book value of total assets as of the end of year t-1 and t. Gr is the firm s sales growth as of the end of year t, equal to the change in sale revenues between current year and previous year divided by total sale revenues of previous year, indicating a firm s investment opportunies. Board is the board size, as measured by the total number of directors on the board. Dir is the proportion of non-executive directors on the board. Dul is a dummy variable taking the value of 1 if the roles of the chief executive officer and board chairman are overlapped, and 0 otherwise. ShortDebt is year-end total current liabilies divided by year-end total debt, reflecting a company s debt matury structure. Bank is bank loans acquired from banks scaled by total debt, which is used to measure debt sources. Lev is the sum of the year-end book value of short-term debt and long-term debt deflated by the year-end book value of total assets. Roa is return on assets as of year t-1, equal to the ratio of the prof before interest and tax to the book value of total assets. Age is the number of years the company has been listed on the stock exchanges in China since IPO.

19 Corporate Governance Reduce the Overinvestment of Free Cash Flow in China? 115 Table 4: Descriptive statistics for variables used to estimate model (3) and (4) Variable No. of obs Mean Median Min Max Std OverI FCF Gr Roa Board Dir Dul ShortDebt Bank Lev Table 5 reports univariate differences in means for all regression variables used in model (3) and (4) between state-controlled companies and private-controlled companies. The sample consists of more state-controlled companies than private-controlled companies, since, as previously mentioned, among the 1411 firm-year observations during the period , only 276 of them are private-controlled companies. The distribution of sample above suggests that state-controlled companies are more likely than private-controlled companies to engage in overinvestment when their free cash flow is high. Consistent wh theoretical expectation, on average, state-controlled companies tend to have a higher overinvestment and free cash flow. The mean values of overinvestment and free cash flow for state-controlled companies are respectively and 0.063, while the corresponding statistics for private-controlled companies are and 0.055, respectively. The diferences in overinvestment and free cash flow between state-controlled companies and private-controlled companies are both statistically significant at conventional levels (wh state-controlled companies exhibing higher mean values than private-controlled companies). Not surprisingly, state-controlled companies have a larger size of the board, more debt-to-asset ratio, lower profabily, and longer time listed on stock exchanges. However, private-controlled companies have a less seperation of the roles of board chairman and CEO, higher proportion of non-executive directors on the board, and higher percentage of short-term debt in total debt. The differences in means of variables above between state-controlled companies and private-controlled companies are all statistically significant. However, there are no significant differences in bank loans and growth opportunies between state-controlled companies and private-controlled companies. In a nutshell, though the univariate analyses above provide strong preliminary supports to the hypotheses developed in this paper, they only show binary correlations whout controlling for other potential determinants. In the next section, I attempt to extend my analysis by more rigorously examining whether the evidence on these hypotheses holds in a multivariate regression framework. This table reports the results for univariate tests for all regression variables between statecontrolled and private-controlled subsample. Among 1411 firm-year observations, statedcontrolled companies and private-controlled companies represent 1135 and 276 of observations, repectively. For each variable, I separately report the variable means for stated-controlled and private-controlled companies, differences in means, t-statistics, and p-values. A company is treated as state-controlled one if s ultimate controlling shareholder is the governments at all levels, such as the bureaus of state assets

20 116 Ji-fu Cai management, finance bureaus and bureaus in charge of different industries or other government agencies et al., and private-controlled companies otherwise. For the definions of OverI, FCF, Board, Dir, Dul, ShortDebt, Bank, Lev, Gr, Roa, and Age see the note to Table 4. 1 Table 5: Univariate tests by the identy of controlling shareholder Variable Stated-controlled Privatecontrolled Differences t-statistics p-values companies (1135) companies (276) in means OverI FCF Board Dir Dul ShortDebt Bank Lev Gr Roa Age Analysis of corelation coefficients Table 6 reports both Spearman and Pearson correlation coefficients between the regression variables used in regression model (3) and (4) which allow for industry and firm level clustering. Spearman (Pearson) correlation coefficients are presented above (below) the main diagonal. In terms of Spearman correlation matrix, It is noted that the ovinvestment variable (OverI ) is significantly posively correlated wh free cash flow ( FCF ), growth opportunies ( Gr ), the board size ( Board ), bank loan ( Bank ), prior period s profabily ( Roa 1 ), and significantly negatively associated wh the proportion of the non-exective directors on the board ( Dir ), short-term debt ( ShortDebt ), total leverage ( Lev ), and the natural logarhm of the number of years listed on stock exchanges in China since IPO ( LnAge ), highlighting the importance of explicly controlling for these company s attributes in the multivariate regressions. Finally, I find that the correlation coefficients betweeen independent variables are generally small, wh the highest (lowest) being (-0.353) between Lev and LnAge Roa ( 1 ), which thus reduces my concerns that multicollineary is possible spuriously responsible for the evidence on the hypotheses developed in this paper. The Pearson correlation coefficients between regression variables also show similar characteristics to Spearman correlation coefficients.